Innernet fails with "Decode error occurred: Failed to parse message with type 16"

[odrling]

The interface is still set up correctly by innernet (or at least it seems to work and I hope I'm not missing something, but the hosts are still pingable).

❯ /usr/local/bin/innernet -vv --mtu 1420 up japanet7 --interval 60 --daemon
[*] bringing up interface japanet7.
[D] set address 10.100.0.6/16 on interface japanet7
[D] set interface japanet7 up with mtu 1420
[D] route 10.100.0.6/16 already existed.
[*] fetching state for japanet7 from server...
[D] [ureq::stream] connecting to 10.100.0.2:61820 at 10.100.0.2:61820
[D] [ureq::stream] created stream: Stream(TcpStream { addr: 10.100.0.6:4204, peer: 10.100.0.2:61820, fd: 4 })
[D] [ureq::unit] sending request GET http://10.100.0.2:61820/v1/user/state
[D] [ureq::unit] writing prelude: GET /v1/user/state HTTP/1.1
Host: 10.100.0.2:61820
User-Agent: ureq/2.9.6
Accept: */*
X-Innernet-Server-Key: REDACTED
[D] [ureq::response] Body entirely buffered (length: 2196)
[D] [ureq::pool] adding stream to pool: http|10.100.0.2|61820 -> Stream(TcpStream { addr: 10.100.0.6:4204, peer: 10.100.0.2:61820, fd: 4 })
[D] [ureq::unit] response 200 to GET http://10.100.0.2:61820/v1/user/state
[D] [wireguard_control::backends::kernel] get_by_name: got 1 response message(s) from netlink request
[D] [wireguard_control::backends::kernel] get_by_name: parsed wireguard device japanet7 with 4 peer(s)

[*] updated interface japanet7

[D] [ureq::stream] dropping stream: Stream(TcpStream { addr: 10.100.0.6:4204, peer: 10.100.0.2:61820, fd: 4 })

[E] Decode error occurred: Failed to parse message with type 16

NOTE: this also happens with no options set in the command line, that's just the command I've taken straight from my service file.

innernet show also fails with just this error:

❯ innernet -vv show

[E] Decode error occurred: Failed to parse message with type 16

innernet fetch doesn't consider the interface to be up.

❯ innernet fetch japanet7

[E] Interface is not up. Use 'innernet up japanet7' instead

The issue seems to have appeared on my server after a kernel update (a vendored SBC kernel 5.10 → 6.8 mainline) and a friend also got this error after an update to Alpine Linux 3.19 (so it would be 6.1 → 6.6).

❯ innernet --version
innernet 1.6.1

[odrling]

I've now looked a bit more into it. This happens when calling get_local_addrs and the error comes from https://github.com/rust-netlink/netlink-packet-core/blob/91e71b69fe8d94a8ae7e2748b0443272c6c3307c/src/message.rs#L116. It seems related to k3s as this doesn't happen when I stop it on my server.

[strohel]

@odrling thanks for the investigation. What is k3s?

[odrling]

It's a Kubernetes implementation https://github.com/k3s-io/k3s

[strohel]

Interesting. Could it be affecting kernel's netlink responses somehow? Maybe they are then extended with some virtualization/namespace/cgroup info then?

[Santonclause]

I experience the same issue and it started after k3s installation. Stopping k3s service does not help

[odrling]

After looking more into it, it seems to be related to this upstream issue https://github.com/rust-netlink/netlink-packet-route/issues/54

Upgrading to 0.18+ probably would fix this but they changed the API with 0.18 it seems, so it's not just a flip of a switch.